Thermoworks TX-1001X-OP & TX-1003X-AP Probe Steinhart-Hart Coefficients


 
T

Tim P

New member
Hi all,

First time poster but I wanted to share some recent information I collected on the Thermoworks probes.

I am not actually a Heatermeter user but given the recommendations here I selected some Thermoworks probes (TX-1001X-OP & TX-1003-AP) for a custom RPi3 build I have been working on. I noticed in an earlier thread (https://tvwbb.com/showthread.php?66718-Thermoworks-1003X-AP) that others had noticed a required offset in some recent implementations. I therefore decided to take the probes in where I have access to a Hart Scientific 24590T dry block Temperature calibration system that is maintained/calibrated against a NIST traceable standard.

z37qzk1.jpg


Using this system, resistance was measured from 0 - 200°C in 25°C increments. Despite presumably using the same internal thermistor, I did find the two probes to respond slightly differently. There could be some manufacturing variability or it is possible that my measurement setup generates this. For example, the narrow barrel diameter of the TX-1001X didn't fit really snug in the calibration block.

CPxig4Q.jpg



I did a fit of the Steinhart-Hart coefficients using sum of least squared error in Excel and got the following results.

[table="width: 500, align: left"]
[tr]
[td]Coefficient[/td]
[td]TX-1001X-OP / TX-1003X-AP[/td]
[/tr]
[tr]
[td]Steinhart A[/td]
[td]0.0007343140544[/td]
[/tr]
[tr]
[td]Steinhart B[/td]
[td]0.0002157437229[/td]
[/tr]
[tr]
[td]Steinhart C[/td]
[td]0.0000000951568577[/td]
[/tr]
[/table]

Finally, I followed this up last night by doing a calibration check with boiling water at home after loading these coefficients into my build code. The plot below also includes the retrospectively back-calculated temperature result that would have been obtained if using the current HeaterMeter Steinhart coefficients. As the earlier thread linked above indicated, these coefficients appear to give results that are a few degrees lower.

oqbFbNp.jpg


Thank you to the community for this great project. Some of the issues that have been encountered and tackled in the hardware here gave me hints on how to approach my own custom project.

Regards,
Tim

Edit: Can anyone give me a hint on why the inline images aren't loading?---FIXED---

Edit2: After further playing around with this, I would just use the coefficients for the TX-1003X-AP probe that I measured. The sum of squared error is two orders of magnitude lower on this profile fit. Today, I put in measured values for the resistors in my voltage divider circuit and got very good agreement between both probes with this correction. For this reason, I am removing the coefficients measured on the TX-1001X-OP probe.
 
Last edited:
Yeah that certainly is interesting. I have never seen a calibration device before so thanks for sharing that!

Interestingly enough, if I plug your coefficients into my comparison tool, I see that your boiling point should be at an ADC reading of 231 = 209.5F. An ADC reading with the current Thermoworks probe coefficients in the snapshot would give a result of 209.8F. 1 LSB of ADC at this range is 0.33F so the result we should be coming up with is within 1 LSB of your results. What version of the OpenWrt firmware are you using? Using the older coefficients you'd get 207.4F which seems closer to what you're getting. If you have the latest coefficients, if you set them to TX-1001X-AP (and save them) then change it to "Custom" the C coefficient should be 1.3193731e-7 for the new version or 9.3692565e-08 for the old.
 
Good call. Out of laziness, I had grabbed the Heatermeter Coefficients from the javascript code in your online comparison tool and it looks like those aren't the same as the values in the current Heatermeter OpenWRT code.

G2dJpDI.jpg


To keep the ADC conversion out of the analysis, I evaluated the current Heatermeter coefficients by predicting temperature values using the Steinhart-hart equation and got much better agreement with the values I generated. The two curves differ a little more below room temperature but no one is bbq'ing at those temperatures. The largest squared error difference arises there and i flagged those cells yellow for emphasis.

3qO1Z1s.jpg


2K5Qra4.jpg


In summary, across the temperature range of use, our two fits very closely agree. However, if one is using the Heatermeter coefficients and checking the calibration with icewater, you will probably observe larger offsets.
 
The "Thermoworks TX-1000 20160910" values from the comparison tool should be the same as those used in the current firmware's snapshot + or - the round-trip rounding that comes from the fact that HeaterMeter only does single precision floating point. I'm happy to see that our numbers are pretty close though, apart from the lower temps. I only calibrated down to around 35C due to my methodology so there's the reason right there probably.

That said, your method is far superior to my own despite the lower sample count so I'll be updating the webui with your coefficients. Thanks for providing them!
 
Bryan Mayland said:
The "Thermoworks TX-1000 20160910" values from the comparison tool should be the same as those used in the current firmware's snapshot + or - the round-trip rounding that comes from the fact that HeaterMeter only does single precision floating point. I'm happy to see that our numbers are pretty close though, apart from the lower temps. I only calibrated down to around 35C due to my methodology so there's the reason right there probably.

That said, your method is far superior to my own despite the lower sample count so I'll be updating the webui with your coefficients. Thanks for providing them!
Click to expand...

Bryan-

How do we get these newest coefficients loaded into our unit (i.e. defaults selected when you pick the thermoworks probe?). Is there a new software/firmware to load? Available from the GUI flash page?
 
No just go to the config page in the webui and select "Custom" as the thermistor type. Paste this in the "A" box 0.0007343140544,0.0002157437229,0.0000000951568577 and hit save. Optionally you could cut that into three pieces and put them into A B and C but who has time for that?
 
Yup, it is already in but won't be published until I get this snapshot working again. Somewhere between the next 24 hours and a couple of days!
 
I just wanted to say that this is out now (in the snapshot) and thanks to Tim for doing the calibration!

I actually just threw a big roast on my grill and was upset because they read 32-33F in the middle of the chunk of meat. I immediately of course blamed the new coefficients for being off when the ChefAlarm was reading 36-37F. Then I realized this was my regular HeaterMeter which didn't have the new snapshot on it so I quickly popped these coefficients in and BAM right to 36F. Great work, Tim.
 
This is great, I only just bought my first thermoworks probe the other week (I know, I'm a slow adapter on thermometer probes - I cook mostly by feel). Upon boiling water calibration they are TONS faster than maverick probes. Maverick probes NO MORE!
 
You must log in or register to reply here.

 

Back
Top